Prediction of Concrete Structure Service Life Based on the Principle of Neural Connections in Brain Circuits

Abstract

Concrete is the most widely used structural material in the world; however, its structural safety is affected by concrete carbonization and reinforcement corrosion. According to the principles and characteristics of concrete carbonization and reinforcement corrosion, this paper establishes artificial neural network prediction and evaluation models for the depth of carbonization and the degree of reinforcement corrosion based on the principle of neural connections in brain circuits. The results show that the artificial neural network prediction models are of high accuracy, and that the RBF neural network prediction model is more accurate and requires far less training time than the BP neural network model.

Deep learning-based prediction for time-dependent chloride penetration in concrete exposed to coastal environment

“…After parameter optimization, values of RSME, MAE, MAPE and R 2 of BP neural network change from 0.0799, 0.0629, 13.40% and 0.9344 to 0.0578, 0.0433, 8.6% and 0.9539, respectively. However, as mentioned above, the models trained by BP and SVM models still need extra input data when predicting the value at a certain time, and the prediction accuracy decreases rapidly with the increase of the time steps [ 38 , 39 ]. In addition, existing researches indicate that, if the sample size and sample feature dimension increase, that is, with the increase of the complexity of the situation, pure mathematical models or prediction model taking traditional machine learning algorithm as the main support would hardly satisfy the practical demands [ 15 , 40 ].…”